Cooperative diagnosis of web transaction failures

1. A method of analyzing Web transaction failures in a network including a plurality of clients, comprising: monitoring, by a client computing device, Web transactions at the client computing device to obtain Web transaction information; recording, by the client computing device, the Web transaction information in a repository that is available for recording of Web transaction information by the plurality of clients; querying, by the client computing device, the repository to obtain selected Web transaction information recorded by some or all of the plurality of clients; and analyzing, by the client computing device, the selected Web transaction information recorded by some or all of the plurality of clients to determine at least one suspect network entity that caused a Web transaction failure, wherein analyzing the selected Web transaction information comprises: for each network entity involved in the Web transaction failure, retrieving a success count and a failure count; calculating a blame score for each network entity involved in the Web transaction failure as a ratio of the failure count to a sum of the success count and the failure count; determining if the blame score for each network entity involved in the Web transaction failure exceeds a first value determined by statistical analysis of blame scores of other network entities; if the blame score for a network entity exceeds said first value, adding the network entity to a set of suspect network entities that caused the Web transaction failure; and for at least selected pairs of network entities involved in the Web transaction failure, retrieving a success count and a failure count for a pair of network entities, calculating a blame score as a ratio of the failure count to the sum of the success count and the failure count for each pair of network entities involved in the Web transaction failure, determining if the blame score for each pair of network entities involved in the Web transaction failure exceeds a second value determined by statistical analysis of blame scores of other pairs of network entities, and if the blame score for the pair of network entities exceeds said second value, adding the pair of network entities to the set of suspects that caused the Web transaction failure.

2. A method as defined in claim 1 , wherein recording the Web transaction information comprises updating success and failure counts for each network entity involved in a Web transaction.

3. A method as defined in claim 1 , further comprising constructing a record for each Web transaction, the record containing the Web transaction information.

4. A method as defined in claim 1 , further comprising performing active measurements by automatically accessing a selected website and recording in the repository Web transaction information related to the selected website.

5. A method as defined in claim 1 , further comprising recording the Web transaction information in a local storage device associated with the client computing device.

6. A method as defined in claim 1 , wherein determining if the blame score for each network entity exceeds said first value comprises retrieving a distribution of blame scores for all network entities of the same functional type as the network entity, and comparing the blame score with the blame score distribution.

7. A method as defined in claim 1 , further comprising determining that a first suspect in the set of suspects is contained in a second suspect in the set of suspects, and removing the first suspect from the set of suspect network entities.

8. A computer-readable storage device encoded with instructions that, when executed by a computing device, perform a method as defined in claim 1 .

9. A method for diagnosing a cause of a Web transaction failure, comprising: for each network entity involved in the Web transaction failure, retrieving, by a client computing device, a success count and a failure count; calculating, by the client computing device, a blame score as a ratio of the failure count to the sum of the success count and the failure count for each network entity involved in the Web transaction failure; determining, by the client computing device, if the blame score for each network entity involved in the Web transaction failure exceeds a first value determined by statistical analysis of blame scores of other network entities; if the blame score for a network entity exceeds said first value, adding, by the client computing device, the network entity to a set of suspects that caused the Web transaction failure; and for at least selected pairs of network entities involved in the Web transaction failure, retrieving a success count and a failure count for a pair of network entities, calculating a blame score as a ratio of the failure count to the sum of the success count and the failure count for each pair of network entities involved in the Web transaction failure, determining if the blame score for each pair of network entities involved in the Web transaction failure exceeds a second value determined by statistical analysis of blame scores of other pairs of network entities, and if the blame score for the pair of network entities exceeds said second value, adding the pair of network entities to the set of suspects that caused the Web transaction failure.

10. A method as defined in claim 9 , wherein determining if the blame score for each network entity exceeds said first value comprises retrieving a distribution of blame scores for all network entities of the same functional type as the network entity, and comparing the blame score with the blame score distribution.

11. A method as defined in claim 9 , further comprising determining that a first suspect in the set of suspects is contained in a second suspect in the set of suspects, and removing the first suspect from the set of suspect network entities.

12. A computer-readable storage device encoded with computer-readable instructions that, when executed by a computing device, perform a method as defined in claim 9 .

13. A computing device comprising: a processor and a storage device encoded with instructions that, when executed by the processor, are configured to: perform Web transactions in response to user requests; monitor the Web transactions and to obtain Web transaction information representative of the Web transactions; record the Web transaction information in a repository that is available for recording of Web transaction information by a plurality of Web clients; and query the repository to obtain selected Web transaction information recorded by some or all of the plurality of Web clients and to analyze the selected Web transaction information to determine at least one suspect network entity that caused a Web transaction failure, including analyzing the selected Web transaction information by: for each network entity involved in the Web transaction failure, retrieving a success count and a failure count; calculating a blame score for each network entity involved in the Web transaction failure as a ratio of the failure count to a sum of the success count and the failure count; determining if the blame score for each network entity involved in the Web transaction failure exceeds a first value determined by statistical analysis of blame scores of other network entities; if the blame score for a network entity exceeds said first value, adding the network entity to a set of suspect network entities that caused the Web transaction failure; and for at least selected pairs of network entities involved in the Web transaction failure, retrieving a success count and a failure count for a pair of network entities, calculating a blame score as a ratio of the failure count to the sum of the success count and the failure count for each pair of network entities involved in the Web transaction failure, determining if the blame score for each pair of network entities involved in the Web transaction failure exceeds a second value determined by statistical analysis of blame scores of other pairs of network entities, and if the blame score for the pair of network entities exceeds said second value, adding the pair of network entities to the set of suspects that caused the Web transaction failure.

14. A computing device as defined in claim 13 , wherein determining if the blame score for each network entity exceeds said first value comprises retrieving a distribution of blame scores for all network entities of the same functional type as the network entity, and comparing the blame score with the blame score distribution.